Statistical Modeling

Understanding Ecological Thresholds in Aquatic Systems Through Retrospective Analysis


Contaminants discharged into aquatic environments are often rapidly scavenged by particles and removed to the sediments. This can lead to the build up of toxic levels of contaminants in sediments leading to adverse ecological effects. However, there are many ways to measure the adverse biological impacts of these contaminants, ranging from short-term ex-situ acute toxicity testing to in-situ assessments of impacts on biological diversity. Although a vast quantity of data has been collected and archived relating the presence of chemical contaminants to a particular biological outcome, there has been limited meta-analyses examining the effects of chemical across a range of outcomes. Through analysis of sediment quality databases such as the USEPA’s National Sediment Quality Survey Database, we will model the effects of sediment contaminants across a wide range of outcomes and along different levels of biological organization, ranging from biomarker expression expressed at the cellular level, to population level effects on diversity.

Experimental Approach

The experimental approach will involve statistical modeling of available data on sediment chemistry with matched toxicity. We propose the use of a two-stage approach to achieve this aim. In the first stage of the analysis, we will estimate associations between chemical concentrations and sediment toxicity and for each toxicity outcome separately. Because interest focuses on the independent effects of multiple chemical concentrations, some of which may be highly collinear, we will use a formal version of principle components regression, known as structural equations modeling (SEM), to quantify the toxicity of groups of chemicals. The resulting estimated coefficients from this SEM analysis, which will represent associations between toxicity and each chemical group, will then be standardized, so that associations among different chemicals and different outcomes are comparable. In the second stage, we will investigate the distribution of the standardized regression coefficients for different toxicity tests, and factors that explain variation in these estimates.

Expected Benefits

It is clear that in order to protect the integrity of aquatic ecosystems, we must have thresholds for allowable levels of contaminants in sediments to determine the presence (or absence) of unacceptable risks of impairment. Given the large number of ways to assess impairment, however, environmental decision makers require guidance on how to incorporate available information to more holistically assess impacts occurring over a wide range of time scales and across varying levels of biological organization. This research will allow decision makers to do just that – to simultaneously incorporate multiple endpoints in assessing allowable levels of contaminants in sediments.